A naukri.com initiative
Netflixtechblog
4w
324
Image Credit: Netflixtechblog
Title Launch Observability at Netflix Scale
- In the blog post 'Launch Observability at Netflix Scale,' Varun Khaitan discusses the strategies and architecture implemented to achieve comprehensive title observability at scale.
- The introduction of observability endpoints was a key step, with each microservice in the Personalization stack required to introduce a 'Title Health' endpoint.
- The endpoints were designed to accurately reflect production behavior, standardize communication, and follow the Insight Triad principle of 'Healthy,' 'why not healthy,' and 'how to fix it.'
- Standardization was achieved through a stable proto request/response format, enhancing adoption, system simplicity, and debuggability for engineers.
- The importance of providing detailed information in endpoint responses to aid Launch Managers and partner engineers in understanding and addressing issues was emphasized.
- A high-level architecture was outlined, detailing the establishment of observability endpoints, proactive monitoring, real-time data tracking, optimized data storage, and APIs for stakeholders.
- Proactive monitoring was conducted through scheduled collector jobs, ensuring title health evaluations for various Netflix rows.
- Real-time Title Impressions were monitored via Kafka Queue, aggregating impressions data to assess title performance in near-real-time.
- Data storage and distribution were facilitated through Hollow Feeds, allowing for efficient dissemination of health data across service boxes.
- An Observability Dashboard, powered by the Health Check Engine, provided stakeholders with current title status across supported rows.
Read Full Article
19 Likes
For uninterrupted reading, download the app